Skip to main content
. Author manuscript; available in PMC: 2021 Sep 1.
Published in final edited form as: Med Phys. 2020 Jun 23;47(9):4125–4136. doi: 10.1002/mp.14308

Fig 3.

Fig 3.

Workflow for external validation. Features extracted from the Cohort-2 are used to externally validate the model trained with no-harmonized features of Cohort-1 (Scenario A). External-validation harmonized features, obtained after Combat application, are used to both models: scenario B refers to the external validation performed with the model trained with no-harmonized features, whereas scenario C represents the external validation of the model trained with harmonized features coming from Cohort-1. Feature extraction (blue box) made use of publicly available tools and was common to both training and external-validation data across cohorts. Model definition and training (yellow box) are described in Fig.2. The green boxes represent the external validation with the three different scenarios. The solid lines follow the training set path, while dashed lines track the external validation process.